3D reverse-time migration using the acoustic wave equation: An experience with the SEG/EAGE data set

Abstract

Kirchhoff is the most commonly used 3D prestack migration algorithm because of its speed and other economic advantages, but it uses a high-frequency ray approximation to the wave equation and, therefore, has difficulties in imaging complex geologic structures where multipathing occurs (e.g., beneath rugose horizons such as faulted salt domes where traveltime calculations become difficult). In contrast to Kirchhoff migration, reverse-time migration computes numerical solutions to the complete wave equation and, therefore, is potentially more accurate. But this technique is not popular in the industry because it is computationally intensive and expensive. However, we felt some recent developments would allow 3D reverse-time migration to be done relatively inexpensively on PC-based distributed memory clusters. In order to examine this hypothesis, we implemented reverse-time migration on a PC cluster by using higher-order accuracy finite difference algorithms (Wu et al., 1996), an excitation time approach (Chang and McMechan, 1994), and variable grids (Mufti et al., 1996) to reduce memory and CPU time. We then used the pseudo-spectral method—following Gazdag (1981), Kosloff and Baysal (1982), and Fornberg (1987)—to further reduce CPU time and core memory requirements. In this article, we compare our reverse-time migration images with first-arrival Kirchhoff migration images to demonstrate that 3D reverse-time migration can produce high fidelity images under the PC-based distributed memory cluster machine.